dates <- c(
    "01/02/03",           # Ambiguous MM/DD/YY
    "01-02-2003",         # DD-MM-YYYY or MM-DD-YYYY
    "January 2, 2003",    # Month name format
    "2003.01.02"          # Period-separated
)

strptime(dates[1], "%m/%d/%y")

strptime(dates[2], "%d-%m-%Y")

strptime(dates[3], "%B %d, %Y")

as.Date(dates[1], format="%m/%d/%y")

as.Date(dates[2], format="%d-%m-%Y")

Sys.setlocale("LC_TIME", "C")  # POSIX locale

format(as.Date("2024-01-01"), "%a %b %d %Y")

Sys.setlocale("LC_TIME", "en_US.UTF-8")  # US locale

format(as.Date("2024-01-01"), "%a %b %d %Y")

ct_ny <- as.POSIXct("2024-01-01 12:00:00", tz="America/New_York")
ct_utc <- as.POSIXct("2024-01-01 12:00:00", tz="UTC")
print(difftime(ct_ny, ct_utc))

ambiguous_date <- "03/04/05"

as.Date(ambiguous_date, format="%m/%d/%y")  # US interpretation

as.Date(ambiguous_date, format="%d/%m/%y")  # European interpretation

as.Date(ambiguous_date, format="%y/%m/%d")  # ISO-like interpretation

# A simple data frame with dates 
dates_df <- tibble(
    date = c("2024-01-01", "2024-02-01", "2024-03-01"),
    value = c(1, 2, 3)
)

dates_df

dates_df |> 
  write_csv("datasets/dates.csv")

read_csv("datasets/dates.csv")

dates_df2 <- dates_df |> 
  mutate(date = as_date(date)) # as_date() is part of the lubridate package 

dates_df2

str(co2)

co2

str(DAAG::greatLakes)

DAAG::greatLakes

supplier <- read_table(
  "Supplier     Cost    Error   ErrorSquared
1   9   -1  1
2   8   -2  4
3   9   -1  1
4   12  2   4
5   9   -1  1
6   12  2   4
7   11  1   1
8   7   -3  9
9   13  3   9
10  9   -1  1
11  11  1   1
12  10  0   0 ", col_names = TRUE
)

supplier

supplier |> 
  ggplot(aes(x = Supplier, y = Cost)) +
  geom_line() +
  geom_point() +
  geom_hline(yintercept = mean(supplier$Cost), color = "red") +
  theme_light()

pc_ebt <- read_table(
  "Year     Earnings_M  Mean    Error   Squared_Error
1985    46.163  48.676  -2.513  6.313
1986    46.998  48.676  -1.678  2.814
1987    47.816  48.676  -0.860  0.739
1988    48.311  48.676  -0.365  0.133
1989    48.758  48.676  0.082   0.007
1990    49.164  48.676  0.488   0.239
1991    49.548  48.676  0.872   0.761
1992    48.915  48.676  0.239   0.057
1993    50.315  48.676  1.639   2.688
1994    50.768  48.676  2.092   4.378", col_names = TRUE
)

pc_ebt

pc_ebt |> 
  ggplot(aes(x = Year, y = Earnings_M)) +
  geom_line() +
  geom_point() +
  geom_hline(yintercept = mean(pc_ebt$Earnings_M), color = "red") +
  theme_light()

quarterly_sales <- read_table(
  "Year     Quarter     Period     Sales
90     1      1     362
90     2      2     385
90     3      3     432
90     4      4     341
91     1      5     382
91     2      6     409
91     3      7     498
91     4      8     387
92     1      9     473
92     2     10     513
92     3     11     582
92     4     12     474
93     1     13     544
93     2     14     582
93     3     15     681
93     4     16     557
94     1     17     628
94     2     18     707
94     3     19     773
94     4     20     592
95     1     21     627
95     2     22     725
95     3     23     854
95     4     24     661", col_names = TRUE
)

quarterly_sales

quarterly_sales |> 
  ggplot(aes(x = Period, y = Sales)) +
  geom_line() +
  geom_point() +
  geom_hline(yintercept = mean(quarterly_sales$Sales), color = "red") +
  theme_light()

quarterly_sales_moving_avg <- quarterly_sales |> 
  mutate(
    moving_avg_3 = slider::slide_dbl(Sales, mean, .before = 1, .after = 1, .complete = TRUE),
    moving_avg_5 = slider::slide_dbl(Sales, mean, .before = 2, .after = 2, .complete = TRUE),
    moving_avg_7 = slider::slide_dbl(Sales, mean, .before = 3, .after = 3, .complete = TRUE)
  )

quarterly_sales_moving_avg

quarterly_sales_moving_avg |> 
  ggplot(aes(x = Period)) +
  geom_line(aes(y = Sales, color = "Original")) +
  geom_line(aes(y = moving_avg_3, color = "3-Period MA")) +
  geom_line(aes(y = moving_avg_5, color = "5-Period MA")) +
  geom_line(aes(y = moving_avg_7, color = "7-Period MA")) +
  geom_smooth(aes(y = Sales, color = "Trend"), 
              method = "lm", formula = y ~ x,
              se = FALSE, linetype = "dashed") +
  scale_color_manual(values = c(
    "Original" = "black",
    "3-Period MA" = "red",
    "5-Period MA" = "blue",
    "7-Period MA" = "green",
    "Trend" = "purple"
  )) +
  labs(color = "Series") +  # Legend title
  theme_light()

# Plot of alpha (1 - alpha)^t for different values of alpha at t = 1:5
alpha_values <- c(0.1, 0.3, 0.5, 0.7, 0.9)
t_values <- 1:5

alpha_table <- tibble()

for (alpha in alpha_values) {
  alpha_table <- alpha_table |> 
    bind_rows(
      tibble(
        alpha = alpha,
        t = t_values,
        value = (1 - alpha)^t
      )
    )
}

alpha_table |>
  ggplot(aes(x = t, y = value, color = factor(alpha))) +
  geom_point() +
  geom_line() +
  labs(
    x = "t",
    y = expression((1 - alpha)^t),
    color = "alpha"
  ) +
  scale_y_continuous(breaks = seq(0, 1, 0.1)) +
  theme_light()

sample_data <- read_table(
  "Data     Fit
6.4      NA
5.6     6.4
7.8     6.2
8.8     6.7
11.0    7.3
11.6    8.4
16.7    9.4
15.3    11.6
21.6    12.7
22.4    15.4", col_names = TRUE
)

sample_data |> 
  ggplot() +
  geom_line(aes(x = seq_along(Data), y = Fit), color = "blue", linetype = "dashed") +
  geom_line(aes(x = seq_along(Data), y = Data), color = "black") +
  theme_light()

ses_rjh <- function(x, alpha = 0.3) {
  n <- length(x)
  S_t <- numeric(n)  # Initialize vector with zeros
  
  # Set first value
  S_t[1] <- x[1]
  
  # Calculate smoothed values
  for(t in 2:n) {
    S_t[t] <- alpha * x[t] + (1 - alpha) * S_t[t-1]
  }
  
  return(S_t)
}

ses_rjh(sample_data$Data, alpha = 0.3)

sample_data |> 
  mutate(fit2 = ses_rjh(Data, alpha = 0.3))  |> 
  ggplot(aes(x = seq_along(Data))) +
  geom_line(aes(y = Data), color = "black") +
  # geom_line(aes(y = Fit), color = "red") +
  geom_line(aes(y = fit2), color = "blue", linetype = "dashed") +
  theme_light()

des_rjh <- function(x, alpha = 0.3, beta = 0.3) {
  n <- length(x)
  S_t <- numeric(n)  # Initialize vector with zeros
  b_t <- numeric(n)  # Initialize vector with zeros
  
  # Set first value
  S_t[1] <- x[1]
  b_t[1] <- x[2] - x[1]
  
  # Calculate smoothed values
  for(t in 2:n) {
    S_t[t] <- alpha * x[t] + (1 - alpha) * (S_t[t-1] + b_t[t-1])
    b_t[t] <- beta * (S_t[t] - S_t[t-1]) + (1 - beta) * b_t[t-1]
  }
  
  return(S_t + b_t)
}

des_rjh(sample_data$Data, alpha = 0.3, beta = 0.3)

sample_data |> 
  mutate(fit2 = des_rjh(Data, alpha = 0.3, beta = 0.3))  |> 
  ggplot(aes(x = seq_along(Data))) +
  geom_line(aes(y = Data), color = "black") +
  # geom_line(aes(y = Fit), color = "red") +
  geom_line(aes(y = fit2), color = "blue") +
  theme_light()

tes_rjh <- function(x, alpha = 0.3, b = 0.3, gamma = 0.3, m = 1) {
  n <- length(x)
  S_t <- numeric(n)  # Initialize vector with zeros
  b_t <- numeric(n)  # Initialize vector with zeros
  c_t <- numeric(n)  # Initialize vector with zeros
  
  # Set first value
  S_t[1] <- x[1]
  b_t[1] <- x[2] - x[1]
  c_t[1] <- x[3] - x[2]
  
  # Calculate smoothed values
  for(t in 2:n) {
    S_t[t] <- alpha * x[t] + (1 - alpha) * (S_t[t-1] + b_t[t-1])
    b_t[t] <- b * (S_t[t] - S_t[t-1]) + (1 - b) * b_t[t-1]
    c_t[t] <- gamma * (S_t[t] - S_t[t-1]) + (1 - gamma) * c_t[t-1]
  }
  
  return(S_t + b_t + c_t)
}

tes_rjh(sample_data$Data, alpha = 0.3, b = 0.3, gamma = 0.3)

sample_data |> 
  mutate(fit2 = tes_rjh(Data, alpha = 0.3, b = 0.3, gamma = 0.3))  |> 
  ggplot(aes(x = seq_along(Data))) +
  geom_line(aes(y = Data), color = "black") +
  # geom_line(aes(y = Fit), color = "red") +
  geom_line(aes(y = fit2), color = "blue") +
  theme_light()

quarterly_sales <- read_table(
  "Year     Quarter     Period     Sales
90     1      1     362
90     2      2     385
90     3      3     432
90     4      4     341
91     1      5     382
91     2      6     409
91     3      7     498
91     4      8     387
92     1      9     473
92     2     10     513
92     3     11     582
92     4     12     474
93     1     13     544
93     2     14     582
93     3     15     681
93     4     16     557
94     1     17     628
94     2     18     707
94     3     19     773
94     4     20     592
95     1     21     627
95     2     22     725
95     3     23     854
95     4     24     661", col_names = TRUE
)

quarterly_sales

quarterly_sales |> 
  mutate(fit2 = tes_rjh(Sales, alpha = 0.7, b = 0.3, gamma = 0.9))  |> 
  ggplot(aes(x = seq_along(Sales))) +
  geom_line(aes(y = Sales), color = "black") +
  geom_line(aes(y = fit2), color = "blue") +
  theme_light()

quarterly_sales_ts_simple <- quarterly_sales |>
  select(Sales) |> 
  ts(start = c(1990, 1), frequency = 4) |> 
  as_tsibble()

quarterly_sales_ts_simple

## Multivariate
z <- ts(matrix(rnorm(300), 100, 3), start = c(1961, 1), frequency = 12)
head(z)

tbl3 <- tibble(
  mth = rep(yearmonth("2010 Jan") + 0:8, each = 3),
  xyz = rep(c("x", "y", "z"), each = 9),
  abc = rep(letters[1:3], times = 9),
  value = rnorm(27)
)
as_tsibble(tbl3, key = c(xyz, abc))

quarterly_sales

quarterly_sales_tsibble <- quarterly_sales |>
  mutate(
    # Convert 2-digit year to 4-digit year
    Year = 1900 + Year,
    # Create proper quarter dates by converting to months (Q1=1, Q2=4, Q3=7, Q4=10)
    Month = (Quarter - 1) * 3 + 1,
    # Create the yearquarter date
    YearQuarter = yearquarter(paste(Year, Month, "01", sep = "-"))
  ) |>
  # Select needed columns and convert to tsibble
  select(YearQuarter, Sales) |>
  as_tsibble(
    index = YearQuarter
  )

quarterly_sales_tsibble

quarterly_sales_tsibble2 <- quarterly_sales |>
 select(Sales) |> 
  ts(start = c(1990, 1), frequency = 4) |>
  as_tsibble()

quarterly_sales_tsibble2

fit_ets <- quarterly_sales_tsibble |>
  model(model_ets = ETS(Sales ~ error("A") + trend("A") + season("A")))

autoplot(quarterly_sales_tsibble, Sales) +
  geom_line(aes(y = .fitted), data = augment(fit_ets), color = "blue")

fit_ets |> 
  forecast(h = 8) |> 
  autoplot(quarterly_sales_tsibble)

fit_ets |> 
  forecast(h = 8) |> 
  autoplot(quarterly_sales_tsibble) +
  geom_line(aes(y = .fitted), data = augment(fit_ets), color = "blue")

msi_historical <- read_tsv("datasets/MSI.txt", col_names = c("year", "Q1", "Q2", "Q3", "Q4"))
msi_historical

msi_ts <- msi_historical |> 
  pivot_longer(cols = -year, names_to = "quarter", values_to = "quarter_value", ) |> 
  separate_wider_delim(cols = quarter_value, names = c("quarter_2", "msi"), delim = " ") |> 
  mutate(msi = as.numeric(str_remove(msi, ","))) |> 
  select(-quarter_2) |> 
  mutate(year_quarter = str_c(year, " ", quarter)) |> 
  select(year_quarter, msi) |> 
  mutate(year_quarter = yearquarter(year_quarter)) |> 
  filter(!is.na(msi)) |>
  as_tsibble(index = year_quarter)

msi_ts

msi_ts |> 
  autoplot(msi)

fit_ets_msi <- msi_ts |> 
  model(model_ets = ETS(msi ~ error("A") + trend("A") + season("A")))

fit_ets_msi |>
  forecast(h = 12) |>
  autoplot(msi_ts) +
  geom_line(aes(y = .fitted), data = augment(fit_ets_msi), color = "blue")

fit_arima_msi <- msi_ts |> 
  model(model_arima = ARIMA(msi, stepwise = FALSE, approximation = FALSE))

fit_arima_msi |>
  forecast(h = 12) |>
  autoplot(msi_ts) +
  geom_line(aes(y = .fitted), data = augment(fit_arima_msi), color = "blue")

msi_three_year_forecast <- fit_arima_msi |>
  forecast(h = 14)

msi_three_year_forecast_annual <- as_tibble(msi_three_year_forecast) |> 
  mutate(year = year(year_quarter)) |> 
  group_by(year) |> 
  summarise(
    forecasted_msi = sum(.mean))

msi_three_year_forecast_annual

msi_long_year <- msi_historical |> 
  pivot_longer(cols = -year, names_to = "quarter", values_to = "quarter_value", ) |> 
  separate_wider_delim(cols = quarter_value, names = c("quarter_2", "msi"), delim = " ") |> 
  mutate(msi = as.numeric(str_remove(msi, ","))) |> 
  select(-quarter_2) |> 
  group_by(year) |>
  summarise(
    msi = sum(msi, na.rm = TRUE)
  ) |>
  ungroup()

msi_long_year

msi_long_year_join <- msi_long_year |> 
  full_join(msi_three_year_forecast_annual, by = "year") |> 
  pivot_longer(cols = c(msi, forecasted_msi), names_to = "type", values_to = "msi")

msi_long_year_join

semi_forecast <- tibble(
  year = 2024:2027,
  semi_foreast_msi = c(12174, 13328, 14507, 15413)
)

semi_forecast

msi_long_year_join |> 
  filter(year >= 2022) |>
  ggplot(aes(x = year)) +
  geom_col(aes(y = msi, fill = type)) +
  geom_point(aes(x = year, y = semi_foreast_msi), data = semi_forecast, color = "black") +
  theme_light()

co2

co2_tsibble <- co2 |> 
  as_tsibble(index = yearmonth)

co2_tsibble

fit_ets_co2 <- co2_tsibble |> 
  model(model_ets = ETS(value ~ error("A") + trend("A") + season("A")))

fit_ets_co2 |> 
  forecast(h = 324) |> 
  autoplot(co2_tsibble) +
  geom_point(x = as_date("2024-10-01"), y = 422, color = "red", size = 4) +
  theme_light()

co2_column_names <- c("year", "month", "decimal_date", "monthly_average", "deseasonalized", "n_days", "stdev_days", "uncertainty_mean")

co2_current_data <- read_table("https://gml.noaa.gov/webdata/ccgg/trends/co2/co2_mm_mlo.txt", skip = 52, col_names = co2_column_names)

co2_current_data

co2_current_data_ts <- co2_current_data |> 
  select(monthly_average) |>
  ts(start = c(1959, 1), frequency = 12) |> 
  as_tsibble()

co2_current_data_ts

fit_ets_co2_current <- co2_current_data_ts |> 
  model(model_ets = ETS(value ~ error("A") + trend("A") + season("A")))

fit_ets_co2_current |> 
       forecast(h = 120) |>
  autoplot(filter(co2_current_data_ts, index >= yearmonth("2000-01-1"))) +
  scale_x_yearmonth(date_labels = "%Y") +
  labs(
    x = "Year",
    y = "CO<sub>2</sub> Concentration (ppm)"
  ) +
  theme_light() +
  theme(axis.title.y = ggtext::element_markdown())